Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Scene classification using a hybrid generative/discriminative approach.

Anna Bosch1, Andrew Zisserman, Xavier Muñoz

  • 1Computer Vision and Robotics Group, Universitat de Girona, Campus Montilivi, Avenida Lluís Santaló s/n, Girona, Spain. aboschr@eia.udg.es

IEEE Transactions on Pattern Analysis and Machine Intelligence
|February 16, 2008
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Occupational exposure to bioaerosols during monk parakeet nest removal in urban parks.

Environmental pollution (Barking, Essex : 1987)·2026
Same author

Determination of Second Metabolic Threshold and Maximal Lactate Steady State using Sweat-Inferred Blood Lactate in Active Individuals and Athletes.

Medicine and science in sports and exercise·2026
Same author

The Sound of Water: Inferring Physical Properties from Pouring Liquids.

IEEE transactions on pattern analysis and machine intelligence·2026
Same author

Coexisting Type 2 Inflammatory Conditions in Patients with Asthma Treated with Dupilumab: the RAPID Registry.

Advances in therapy·2026
Same author

Impact of Dexamethasone on Respiratory Sequelae in the Follow-Up of Hospitalized Patients with Sars-Cov-2 Pneumonia.

Infection and drug resistance·2026
Same author

Optimizing GRIDSS for clinical use: A targeted NGS filtering strategy for germline structural variant detection.

European journal of human genetics : EJHG·2026
Same journal

HardFlow: Hard-Constrained Sampling for Flow-Matching Models Via Trajectory Optimization.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Industrial Brain: Self-Evolving Neuro-Symbolic Autonomy with Causal Resilience for Cyber-Physical Systems.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Adaptive Hardness-Driven Dictionary Distillation for Incomplete Streaming View Clustering.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Mixture of Global and Local Experts with Diffusion Transformer for Controllable Face Generation.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Task-KV: Task-aware KV Cache Optimization via Semantic Differentiation of Attention Heads.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Achieving Text-based Person Retrieval with Any Granularity.

IEEE transactions on pattern analysis and machine intelligence·2026
See all related articles

Dimensionality reduction with latent generative models improves weakly supervised scene classification. This approach outperforms standard bag-of-visual-words methods by discovering latent topics for better image categorization.

Area of Science:

  • Computer Vision
  • Machine Learning
  • Pattern Recognition

Background:

  • Weakly supervised scene classification requires categorizing images with limited labels.
  • Traditional methods often rely on direct feature representations like bag-of-visual-words.
  • Latent generative models offer a potential for improved feature extraction.

Purpose of the Study:

  • To evaluate the effectiveness of dimensionality reduction using latent generative models for scene classification.
  • To compare a probabilistic Latent Semantic Analysis (pLSA) based approach against direct bag-of-visual-words representations.
  • To investigate the impact of vocabulary size, topic number, and classifier choice on performance.

Main Methods:

  • Applied probabilistic Latent Semantic Analysis (pLSA) to a bag-of-visual-words representation of images to discover latent topics.

Related Experiment Videos

  • Trained multi-way classifiers on the resulting topic distribution vectors.
  • Introduced a novel vocabulary using dense color SIFT descriptors.
  • Compared performance against direct bag-of-visual-words vectors using k-nearest neighbor and Support Vector Machine (SVM) classifiers.
  • Investigated the effect of spatial information and applied the method to image retrieval and video scene classification.

    • Main Results:

    • The pLSA-based dimensionality reduction approach achieved superior classification performance compared to direct bag-of-visual-words methods.
    • Performance improvements were observed across various configurations of visual vocabulary size, number of latent topics, and classifier types.
    • The study demonstrated the benefit of latent topic discovery for enhancing scene classification accuracy.
    • Spatial information was also found to contribute positively to classification performance.

    Conclusions:

    • Dimensionality reduction using latent generative models, specifically pLSA, is beneficial for weakly supervised scene classification.
    • The proposed method surpasses existing bag-of-visual-words techniques in terms of classification accuracy.
    • The approach shows promise for applications in image retrieval and video scene analysis.